Ovarian Cancer Detection in CT Scan Images Using Transfer Learning

This research paper presents a system for classifying and detecting ovarian CT scan images using transfer learning and examines the effects of different tools on the solution quality. The methodology involves preprocessing, data augmentation, training, transformation, and filtering. The implementation experiments evaluate the system's performance with regard to using different combinations of color modes, filters, and training methods with and without augmentation of the training dataset in addition to the effect of using the discrete wavelet transform (DWT). Analysis of the implementation results indicated that augmentation and the DWT generally improved the results, while the choice of filters and color mode had a limited impact. Furthermore, using the stochastic gradient descent with momentum (SGDM) training method yielded better results compared to the Adam method. Overall, the recommended approach for the best outcomes consists of using the RGB color mode, SGDM training, augmentation, and the DWT without additional filters.